Ekstraksi Tekstur Citra Menggunakan Gaussian dan Multi-Block Local Binary Pattern untuk Identifikasi Tumbuhan Obat

Abstract

Plants identification automatically is still a uses for recognizing various kinds of house plant species and medicinal plants. This research uses a method Multi-Block Local Binary Pattern (MBLBP) descriptor to extract texture feature and Probabilistic Neural Network (PNN) classifying for identifying a house plants and medicinal plants automatically. There are three kinds of MBLBP descriptor used in this research, i.e., , , , , and , . For training and testing, this research uses database of 1440 medicinal plant leaf images and 300 tree images belonging to 30 different types and obtained from Biofarmaka IPB, Cikabayan Farm, Green house Center Ex- Situ Conservation of Medicinal Plant Indonesia Tropical Forest, and Gunung Leutik. The experimental result shows that the concatination of , has the best accuracy in identifying house plants with an accuracy of 77.78%. It shows that MBLBP method is better than LBP method in identifying house plants based on the increase accuracy by 4.45%.